CN106154180A - Energy-storage battery charge/discharge anomaly detection method and detecting system - Google Patents

Abstract

The invention discloses a kind of energy-storage battery charge/discharge anomaly detection method and detecting system.Wherein, the method includes extracting described energy-storage battery electric current under charge/discharge state, voltage data；By the current data and the charge/discharge current threshold ratio that extract relatively, and determine the described charge/discharge state of described energy-storage battery according to comparative result；Utilize K arest neighbors method, determine in described energy-storage battery current/voltage spatial correspondence under described charge/discharge state K nearest neighbor distance and；By K nearest neighbor distance under described charge/discharge state and compare with the distance threshold under charge/discharge state, and detect the charge/discharge Deviant Behavior of described energy-storage battery according to comparative result.By the embodiment of the present invention, can be to find that system aging, fault etc. provide in time to support.

Description

Energy-storage battery charge/discharge anomaly detection method and detecting system

Technical field

The present embodiments relate to battery performance detection technique field, especially relate to a kind of energy-storage battery charge/discharge different Often behavioral value method and detecting system.

Background technology

Generation of electricity by new energy has become as the important component part of China's supply of electric power, and energy-storage battery is as in wind-light storage system Important step, both can be as controlled-load, it is also possible to as controlled source, it is achieved the transmission of the electric energy of charge and discharge.So And, owing to energy-storage battery uses the particularity of environment, and the improper use in practical operation, battery can occur in various degree Aging or fault, the energy-storage battery charge or discharge dystropy brought may affect the even running of whole wind-light storage system And optimum control, cause immeasurable economic loss, even can lead to a disaster.Therefore, system aging, fault are found in time Deng, it is to avoid the peril that excessively use or mistake use cause is the problem that assistant officer is to be solved.

In view of this, the special proposition present invention.

Summary of the invention

For solving technical problem present in prior art, the embodiment of the present invention proposes one and solves this at least in part and ask The energy-storage battery charge/discharge anomaly detection method of topic.In addition, it is also proposed that a kind of energy-storage battery charge/discharge Deviant Behavior inspection Examining system.

To achieve these goals, in the one side of the embodiment of the present invention, techniques below scheme proposed:

A kind of energy-storage battery charge/discharge anomaly detection method, described detection method includes:

Extract described energy-storage battery electric current under charge/discharge state, voltage data；

By the current data of extraction with charge/discharge current threshold ratio relatively, and described energy-storage battery is determined according to comparative result Described charge/discharge state；

Utilize K arest neighbors method, determine described energy-storage battery current-voltage space pair under described charge/discharge state Should be related to middle K nearest neighbor distance and；

By K nearest neighbor distance under described charge/discharge state and compare with the distance threshold under charge/discharge state Relatively, the charge/discharge Deviant Behavior of described energy-storage battery and is detected according to comparative result.

Further, described extraction described energy-storage battery electric current under charge/discharge state, voltage data, specifically include:

Based on training set, build current data rectangular histogram；

Based on described current data rectangular histogram, by CURRENT DISTRIBUTION probability threshold value calculating current frequency threshold value and according to described electricity Stream frequency threshold value determines the off working state of described energy-storage battery；

Characteristic distributions based on current data, rejects electric current occurrence number more than under the described off working state of pre-determined number Electric current, voltage data, the electric current after being screened, voltage data.

Further, described charge/discharge current threshold value determines according in the following manner:

Judge that whether the current data of described extraction is more than zero；

If it is, determine that described energy-storage battery is in described discharge condition, and determine described electric discharge according to below equation Current threshold:

I_Td=min (I_d)

Wherein, described I_dRepresent the current data of described extraction；Described I_TdRepresent described discharging current threshold；

Otherwise, it determines described energy-storage battery is in described charged state, and determine described charging current according to below equation Threshold value:

I_Tc=max (I_c)

Wherein, described I_cRepresent the current data of described extraction；Described I_TcRepresent described charging current threshold value.

Further, the distance threshold under described charge/discharge state determines according in the following manner:

Corresponding according to the described current-voltage space that below equation calculates under described charged state based on training set structure In relation, the often electric current under group trickle charge state, voltage data average in its charged state persistent period section:

$I_{cm}=\frac{1}{s}\underset{i=1}{\overset{s}{\Σ}}{I}_{c,t_i}$

$V_{cm}=\frac{1}{s}\underset{i=1}{\overset{s}{\Σ}}{V}_{c,t_i}$

Wherein, described s represents the charged state persistent period；Described t_iRepresent the moment that charged state is corresponding；Described DescribedRepresent t under current trickle charge state respectively_iThe electric current in moment, voltage data value；Described I_cm, described V_cmRespectively Represent the electric current under described often group trickle charge state, voltage data average in its charged state persistent period section；

The electric current under all continuous discharge states, voltage data is calculated in its discharge condition persistent period according to below equation Average in Duan:

$I_{dm}=\frac{1}{s^{\′}}\underset{i=1}{\overset{s^{\′}}{\Σ}}{I}_{d,t_i}$

$V_{dm}=\frac{1}{s^{\′}}\underset{i=1}{\overset{s^{\′}}{\Σ}}{V}_{d,t_i}$

Wherein, described s' represents the discharge condition persistent period；Described t_iRepresent the moment that described discharge condition is corresponding；DescribedDescribedRepresent t under current continuous discharge state respectively_iThe electric current in moment, voltage data value；Described I_dm, described V_dm Represent the electric current under described all continuous discharge states, the voltage data average in its discharge condition persistent period section respectively；

Electric current, voltage under each described trickle charge, discharge condition is calculated in described training set respectively according to below equation K nearest neighbor distance of data mean value and:

$D_{c,j}=\underset{i=1}{\overset{K}{\Σ}}\left|\right|\left(\begin{array}{c}{I}_{cm,j}\\ V_{cm,j}\end{array}\right)-\left(\begin{array}{c}{I}_{cm,i}^{*}\\ V_{cm,i}^{*}\end{array}\right)|{|}_2^2$

$D_{d,j}=\underset{i=1}{\overset{K}{\Σ}}\left|\right|\left(\begin{array}{c}{I}_{dm,j}\\ V_{dm,j}\end{array}\right)-\left(\begin{array}{c}{I}_{dm,i}^{*}\\ V_{dm,i}^{*}\end{array}\right)|{|}_2^2$

Wherein, described I_cm,jRepresent current data average under jth trickle charge state；Described V_cm,jRepresent described jth Voltage data average under trickle charge state；DescribedDescribedRepresent respectively jth electric current under described charged state, The corresponding electric current of i-th distance, voltage data average in K arest neighbors of voltage data average；Described D_c,jRepresent described jth Electric current, voltage data average I under individual trickle charge state_cm,j、V_cm,jK nearest neighbor distance and；Described I_dm,jRepresent jth Current data average under continuous discharge state；Described V_dm,jRepresent voltage data average under described jth continuous discharge state；Institute StateDescribedRepresent in K arest neighbors of jth electric current under described discharge condition, voltage data average corresponding respectively The electric current of i-th distance, voltage data average；Described D_d,jRepresent electric current, voltage data under described jth continuous discharge state Average I_dm,j、V_dm,jK nearest neighbor distance and；Described K takes odd number；Described j span is by the most continuous in described training set Charging, discharge condition number determine；

Under described charge/discharge state, according to predetermined confidence level α by all described electric currents, the K of voltage data average Individual nearest neighbor distance with distance corresponding to α quantile be defined as the distance threshold under charge/discharge state.

Further, described method also includes:

Obtain the original electric current of described energy-storage battery, voltage data；

Traversal total data, and with spot patch foot abnormal storage point, it is thus achieved that electric current per minute, voltage data；

It is normalized according to below equation:

$I_n=\frac{I}{100}$

V_n=V-3

Wherein, described I represents described current data per minute；Described V represents described voltage data per minute；Described I_nRepresent the current data after normalization；Described V_nRepresent the voltage data after normalization.

Further, described method also includes: carry out parameter initialization, wherein, parameter include obtaining described primary current, K value and confidence level in the natural law of voltage data, training set size, distribution probability threshold value, K arest neighbors method.

To achieve these goals, another aspect according to embodiments of the present invention, it is also proposed that a kind of energy-storage battery charge/discharge Unusual checking system, described detecting system includes:

Extraction module, for extracting described energy-storage battery electric current under charge/discharge state, voltage data；

State determining module, for current data and the charge/discharge current threshold ratio that will extract relatively, and according to comparative result Determine the described charge/discharge state of described energy-storage battery；

Distance and determine module, is used for utilizing K arest neighbors method, determines that described energy-storage battery is in described charge/discharge state Under current-voltage spatial correspondence in K nearest neighbor distance and；

Detection module, for by K nearest neighbor distance under described charge/discharge state and with under charge/discharge state away from Compare from threshold value, and detect the charge/discharge Deviant Behavior of described energy-storage battery according to comparative result.

The embodiment of the present invention proposes a kind of energy-storage battery charge/discharge anomaly detection method and detecting system.Based on from Electric current under the charging and discharging state gathered in energy-storage battery, voltage data, through data prediction, discharge and recharge data screening, Build the current-voltage spatial correspondence of energy-storage battery under charging and discharging state in training set respectively, use and calculate based on K arest neighbors Method calculate under each energy-storage battery charging and discharging state its K nearest neighbor distance and, and obtain distance threshold；Different in actual discharge and recharge In the often behavioral value stage, judge its charging, discharge condition according to current time electric current, voltage data, and calculate corresponding electric current- K nearest neighbor distance and whether exceed distance threshold in voltage space corresponding relation, it is achieved the detection of discharge and recharge Deviant Behavior, tool There are stronger practicality and accuracy, can be to find that system aging, fault etc. provide in time to support.

Accompanying drawing explanation

Fig. 1 is to illustrate according to the energy-storage battery charge/discharge anomaly detection method flow process shown in one embodiment of the invention Figure；

Fig. 2 is according to the current data rectangular histogram schematic diagram in the energy-storage battery training set shown in one embodiment of the invention；

Fig. 3 is according to the abnormality detection schematic diagram based on K nearest neighbor algorithm shown in one embodiment of the invention；

Fig. 4 is that the structure according to the energy-storage battery charge/discharge unusual checking system shown in one embodiment of the invention is shown It is intended to.

Detailed description of the invention

In order to be illustrated more clearly that the object, technical solutions and advantages of the present invention, the present invention each is described in detail below Step, and referring to the drawings and combine instantiation and be described in further detail.

The basic thought of the present invention is that whether electric current, voltage data meet by under monitoring energy-storage battery charging and discharging state Intrinsic rule, to detect whether energy-storage battery discharge and recharge behavior occurs exception.

Electric current from the charging and discharging state that energy-storage battery gathers, voltage data, owing to the packet loss in its transmitting procedure is existing As causing existing more abnormity point, and electric current, voltage data computational accuracy and numerical range are the most inconsistent, it is difficult to directly than Relatively；Energy-storage battery off working state occurs the moment more, it is impossible to direct analysis, needs the work to energy-storage battery/inoperative shape State is divided, and extracts the electric current under charging and discharging state, voltage data；Under charging and discharging state, the Changing Pattern of energy-storage battery Difference, needs to build respectively the unusual checking model of discharge and recharge.

The embodiment of the present invention provides a kind of energy-storage battery charge/discharge anomaly detection method.This detection method includes:

S100: extract energy-storage battery electric current under charge/discharge state, voltage data.

Wherein, this step can specifically include:

S101: based on training set, builds current data rectangular histogram.

Wherein, training set includes current data and voltage data.Per minute electric current, voltage data are sampled once, and all For common energy-storage battery data type.Such as, for one day, 1440 sample points of can sampling；For N days, then training set Upper electric current, voltage data are respectively arranged with the sample point of 1440 × N.

Fig. 2 schematically illustrates the current data rectangular histogram in energy-storage battery training set.Wherein, mark from S201 circle Electric current can be seen that near the 0 point point of null value (current value be), the frequency of non-operating current is the highest, and other current values frequencies Secondary relatively low, as shown in the electric current that S202 triangle marks.

S102: based on current data rectangular histogram, by CURRENT DISTRIBUTION probability threshold value calculating current frequency threshold value and according to electric current Frequency threshold value determines the off working state of energy-storage battery.

This step judges the duty of energy-storage battery based on current data, to determine the inoperative of energy-storage battery State, thus carry out follow-up process.

S103: characteristic distributions based on current data, rejects the electric current occurrence number off working state more than pre-determined number Under electric current, voltage data, the electric current after being screened, voltage data.

Energy-storage battery is in off working state every day mostly.In a non-operative state, current data has and is distributed spy as follows Point: electric current consecutive variations under charging and discharging state, thus can cause 0 neighbouring non-operating current frequency the highest.This step root Electric current, voltage under the electric current occurrence number off working state more than pre-determined number is rejected according to the histogrammic statistical result of current data Data that is to say current convergence off working state data near null value.Such as: in specific implementation process, electricity can be rejected Stream occurrence number is more than electric current, voltage data under the off working state of 1440 γ × n times, and wherein, N represents training set size；γ Represent distribution probability threshold value.

S110: by the current data of extraction with charge/discharge current threshold ratio relatively, and determine energy-storage battery according to comparative result Charge/discharge state.

The embodiment of the present invention can determine discharging current threshold and charging current threshold value according in the following manner.

Assume I_wRepresent the current data after screening.

Judge that whether the current data extracted is more than zero.

If I_w＞ 0, then energy-storage battery is in discharge condition, records moment corresponding to each continuous discharge state and extraction Electric current that this each moment is corresponding and voltage data.If the current data extracted (the i.e. moment pair corresponding to continuous discharge state The current data answered) it is I_d, then determine discharging current threshold I according to below equation_Td:

I_Td=min (I_d)；

If I_w＜ 0, then energy-storage battery is in charged state, records the moment that each trickle charge state is corresponding, and carries Take electric current corresponding to this each moment and voltage data.If the current data extracted (the i.e. moment corresponding to trickle charge state Corresponding current data) it is I_c, then determine charging current threshold value I according to below equation_Tc:

I_Tc=max (I_c)。

Above-mentioned being only assumed as is illustrated, and is not construed as the improper restriction to scope.

Describe in detail with a preferred embodiment below and determine discharging current threshold and the process of charging current threshold value.

This preferred embodiment is with JIUYUE in 2013 certain energy-storage battery data of the 1st～10 minute on the 13rd as training set.

Pretreated current data is 0.5,0.5,0.3 ,-0.1 ,-0.5,0.4,0.5,0.2,0.5 ,-0.3, wherein wraps Containing two continuous discharge states, if discharging time is respectively { 1,2,3} and { 6,7,8,9}, and extract two respectively continuously The electric current inscribed during discharge condition correspondence, voltage data, discharging current threshold is the minima of electric current in all discharging times, i.e. I_Td=0.2；If above-mentioned pretreated packet contains two trickle charge states, be respectively when charging 4,5}, 10}, And inscribe when extracting two trickle charge state correspondences respectively electric current, voltage data, charging current threshold value is all chargings The maximum I of electric current in moment_Tc=-0.1.

S120: utilize K arest neighbors method, determines that energy-storage battery current-voltage space correspondence under charge/discharge state is closed In system K nearest neighbor distance and.

S130: by K nearest neighbor distance under charge/discharge state and enter with the distance threshold under charge and discharge state respectively Row compares, and detects the charge/discharge Deviant Behavior of energy-storage battery according to comparative result.

In this step, distance threshold includes the distance threshold under charged state and the distance threshold under discharge condition.Distance Threshold value can determine in the following manner:

Training set builds the current-voltage spatial correspondence of energy-storage battery under charging and discharging state respectively, and uses Obtain meeting under charge and discharge state the distance threshold of confidence level based on the training of K arest neighbors method.

Specifically, the distance threshold under charge and discharge state can be determined by following steps:

Step A: close according to the current-voltage space correspondence that below equation calculates under charged state based on training set structure In system, the often electric current under group trickle charge state, voltage data average in its charged state persistent period section:

$I_{cm}=\frac{1}{s}\underset{i=1}{\overset{s}{\Σ}}{I}_{c,t_i}$

$V_{cm}=\frac{1}{s}\underset{i=1}{\overset{s}{\Σ}}{V}_{c,t_i}$

Wherein, s represents the charged state persistent period；t_iRepresent the moment that charged state is corresponding；Represent respectively T under current trickle charge state_iThe electric current in moment, voltage data value；I_cm、V_cmRepresent respectively and often organize under trickle charge state Electric current, the voltage data average in its charged state persistent period section.

In this step, in the charge state, s was determined by the actual charged state persistent period.

Step B: calculate the electric current under all continuous discharge states according to below equation, voltage data is held in its discharge condition Average in the continuous time period:

$I_{dm}=\frac{1}{s^{\′}}\underset{i=1}{\overset{s^{\′}}{\Σ}}{I}_{d,t_i}$

$V_{dm}=\frac{1}{s^{\′}}\underset{i=1}{\overset{s^{\′}}{\Σ}}{V}_{d,t_i}$

Wherein, s' represents the discharge condition persistent period；t_iRepresent the moment that discharge condition is corresponding；Table respectively Show t under current continuous discharge state_iThe electric current in moment, voltage data value；I_dm、V_dmRepresent respectively under all continuous discharge states Electric current, voltage data average in its state duration section.

Step C: calculate in training set electric current, voltage number under each trickle charge, discharge condition respectively according to below equation According to average K nearest neighbor distance and:

$D_{c,j}=\underset{i=1}{\overset{K}{\Σ}}\left|\right|\left(\begin{array}{c}{I}_{cm,j}\\ V_{cm,j}\end{array}\right)-\left(\begin{array}{c}{I}_{cm,i}^{*}\\ V_{cm,i}^{*}\end{array}\right)|{|}_2^2$

$D_{d,j}=\underset{i=1}{\overset{K}{\Σ}}\left|\right|\left(\begin{array}{c}{I}_{dm,j}\\ V_{dm,j}\end{array}\right)-\left(\begin{array}{c}{I}_{dm,i}^{*}\\ V_{dm,i}^{*}\end{array}\right)|{|}_2^2$

Wherein, I_cm,jRepresent current data average under jth trickle charge state；V_cm,jRepresent jth trickle charge state Lower voltage data average；Represent jth electric current, K arest neighbors of voltage data average under charged state respectively The middle corresponding electric current of i-th distance, voltage data average；D_c,jRepresent that electric current under jth trickle charge state, voltage data are equal Value I_cm,j、V_cm,jK nearest neighbor distance and；I_dm,jRepresent current data average under jth continuous discharge state；V_dm,jRepresent Voltage data average under jth continuous discharge state；Represent jth electric current under discharge condition, voltage number respectively According to corresponding in K arest neighbors of the average electric current of i-th distance, voltage data average；D_d,jRepresent jth continuous discharge state Lower electric current, voltage data average I_dm,j、V_dm,jK nearest neighbor distance and；K takes odd number；J span is by respective in training set Trickle charge, discharge condition number determine.

In above-mentioned formula, for parameter j, such as, have 10 trickle charge states, 5 continuous discharge states, then fill J under electricity condition is 10 to the maximum, and under discharge condition, j is 5 to the maximum.

Step D: under charge/discharge state, according to predetermined confidence level α by individual for the K of all electric currents, voltage data average Nearest neighbor distance with distance corresponding to α quantile be defined as the distance threshold under charge and discharge state.

Fig. 3 schematically illustrates abnormality detection schematic diagram based on K nearest neighbor algorithm.Wherein, in the charge state, electricity Stream, voltage data average in respective persistent state, and the distribution in current-voltage spatial correspondence, such as S401 The all scatterplot pointed out: under charging normal behavior, K in current-voltage spatial correspondence is individual for electric current, voltage data average Nearest neighbor distance and less, and less than distance threshold T_d1, as shown in the triangle midpoint that S402 points out；In a certain abnormal charging row Under for, K nearest neighbor distance in current-voltage spatial correspondence of electric current, voltage data average and relatively big, and more than away from From threshold value T_d1, as shown in the circle midpoint that S403 points out.In the discharged condition, electric current, voltage data are in respective persistent state Average, and the distribution in current-voltage spatial correspondence, all scatterplot pointed out such as S404: in regular picture behavior Under, K nearest neighbor distance in current-voltage spatial correspondence of electric current, voltage data average and less, and less than distance Threshold value T_d2, as shown in the triangle midpoint that S405 points out；Under a certain paradoxical discharge behavior, electric current, voltage data average are at electricity K nearest neighbor distance of stream-voltage space corresponding relation and relatively big, and more than distance threshold T_d2, as shown in S406 circle midpoint. As seen in Figure 3 under charge/discharge state in current-voltage spatial correspondence, discharge and recharge based on K nearest neighbor algorithm Anomaly detection method can detect the Deviant Behavior in charge and discharge process effectively.

Preferably, the embodiment of the present invention can also include data prediction step.Wherein, in energy-storage battery training set Electric current, voltage data carry out data outliers process and normalized.

Specifically, data outliers processes and may include that

Step a: obtain the original electric current of energy-storage battery, voltage data.

Wherein, for example, it is possible to acquisition precision be 0.1A, numerical range is the current data of-65A～65A and precision is 0.001V, numerical range are the voltage data of 2.5V～3.7V.

Step b: traversal total data, and with spot patch foot abnormal storage point, it is thus achieved that electric current per minute, voltage data.

Normalized may include that and is normalized according to below equation:

$I_n=\frac{I}{100}$

V_n=V-3

Wherein, I represents current data per minute；V represents voltage data per minute；I_nRepresent the electric current after normalization Data；V_nRepresent the voltage data after normalization.

Can be according to the value model of the span of the current data collected and precision and voltage data in this step Enclose with precision to determine normalization mode.

With precision as 0.1A, numerical range be the current data of-65A～65A and precision as 0.001V, numerical range be As a example by the voltage data of 2.5V～3.7V, then the current range after normalization-0.65～0.65, precision be 0.001, normalization After voltage range-0.5～0.7, precision be 0.001.Visible, the electric current after normalization, voltage data have identical meter Calculate precision and close numerical range, thereby may be ensured that what ensuing data calculated is smoothed out.

By data are carried out pretreatment, it is possible to obtain electric current complete, effective, voltage data, thereby may be ensured that electricity Stream, voltage data computational accuracy and numerical range are consistent, and the trouble-free operation for the embodiment of the present invention provides data basis.

Preferably, the embodiment of the present invention can also include parameter initialization step.Wherein it is possible to arrange acquisition electric current, electricity K value and confidence level in the pressure natural law of data, training set size, distribution probability threshold value, K arest neighbors method.

Each parameter in the present embodiment can be adjusted according to practical situation, to meet actual demand.Such as: can basis Relevant parameter required for actual demand definition data calculating, such as (comprises voltage, electric current number with 30 day data of stable operation According to) as training set and arrange parameter.The span of α can be located between (0,1).

The mistake of the charge/discharge unusual checking that the embodiment of the present invention proposes is described below in detail with a preferred embodiment Journey.

Step E: carry out parameter initialization.Wherein, parameter can include but not limited to obtain electric current, the sky of voltage data K value, confidence level in number, training set size, distribution probability threshold value, K arest neighbors method.

Step F: obtain the current electric current of energy-storage battery, voltage data.

Step G: the current electric current of energy-storage battery, voltage data are carried out data outliers process, obtains complete every point Clock electric current, voltage data.

Step H: complete electric current per minute, voltage data are normalized.

Step I: the current data after normalized is compared with charging current threshold value and discharging current threshold respectively Relatively；If the current data after normalized is less than charging current threshold value, then energy-storage battery is in charged state, and performs step Rapid J is to step M；If the current data after normalized is more than discharging current threshold, then energy-storage battery is in discharge condition, And perform step N to step Q.

Step J: calculating current, voltage data average and K nearest neighbor distance in current-voltage spatial correspondence thereof With.

Step K: compare by this distance with the distance threshold under charged state, if this distance and more than charging shape Distance threshold under state, then perform step L；Otherwise, step step M is performed.

Step L: determine energy-storage battery charging dystropy.

Step M: determine that energy-storage battery charging behavior is normal.

Step N: calculating current, voltage data average and K nearest neighbor distance in current-voltage spatial correspondence thereof With.

Step O: compare by this distance with the distance threshold under discharge condition, if this distance and more than electric discharge shape Distance threshold under state, then perform step P；Otherwise, step Q is performed.

Step P: determine energy-storage battery electric discharge dystropy.

Step Q: determine that energy-storage battery electric discharge behavior is normal.

It should be noted that unless stated otherwise, the most identical symbol can represent identical implication.

The process of energy-storage battery unusual checking is described as a example by below using in JIUYUE, 2013 data as training set. Wherein, N=30.

In training set after data prediction, it is thus achieved that electric current I_n, voltage V_n.With computational accuracy 0.001 for group away from structure Electric current I_nRectangular histogram, and according to given distribution probability threshold value (such as: γ=5%), reject electric current occurrence number and be more than Electric current under the off working state of 1440 γ × N=2160 time, voltage data, the electric current I after being screened_w, voltage V_wData. Work as I_wDuring ＞ 0, energy-storage battery is in discharge condition, is changed to cut-point with the discharge condition of energy-storage battery, records continuous discharge The moment that state is corresponding, and extract electric current I corresponding to each moment_d, voltage V_dData, and discharging current threshold is (such as: I_Td= min(I_d)=0.008)；Work as I_wDuring ＜ 0, energy-storage battery is in charged state, is changed to segmentation with the charged state of energy-storage battery Point, record moment corresponding to trickle charge state, and extract electric current I corresponding to each moment_c, voltage V_cData, and charging current Threshold value (I_Tc=max (I_c)=-0.007).Build the space length model of current-voltage relation under charging, discharge condition respectively, And use K arest neighbors method to calculate electric current, (K=3) 3 minimum distances of voltage data average under current charging, discharge condition Sum, chosen distance with in numerical value corresponding to (α=5%) 5% quantile as distance threshold, obtain: T_d1=0.000645, T_d2=0.000494, and realize unusual checking with this.

The electric current of energy-storage battery that collects using on March 12nd, 2015, voltage data are as a example by test set, through data Charging current I after normalization on the same day is obtained after pretreatment_n, voltage V_n.If I_n＜ I_Tc, energy-storage battery is in charged state, and record is worked as Front moment electric current I_c, voltage V_cData；And calculate its average I_cm、V_cmIn current-voltage spatial correspondence K arest neighbors away from From and D_c.If D_c＞ T_d1, energy-storage battery charging dystropy；Otherwise, charging behavior is normal.There are 6 continuous print chargings the same day State, persistent period, electric current, voltage data average, and K in current-voltage spatial correspondence in the charge state Nearest neighbor distance and as shown in Table 1:

Table one:

t	Icm	Vcm	Dc	Testing result
					632～641	-0.0411	0.2921	0.0000441	Normally
672～676	-0.0412	0.3162	0.000155	Normally
					687～691	-0.0572	0.2958	0.0000758	Normally
704～708	-0.0576	0.3060	0.0000282	Normally
					826～832	-0.0917	0.3126	0.000189	Normally
943～996	-0.1356	0.3392	0.0019	Abnormal

Wherein, the 6th charging interval section is the same day the 943rd～996 point, K nearest neighbor distance and be 0.0019, and it is more than Distance threshold (T under charged state_d1=0.000645), thus testing result is charging of corresponding moment dystropy；And remaining time The distance that obtains and respectively less than distance threshold under the charged state carved, then testing result is that charging behavior is normal.

There are 3 lasting discharge conditions the same day on March 12nd, 2015, persistent period, electric current, voltage data average, and Its K nearest neighbor distance and as shown in Table 2 in current-voltage spatial correspondence in the discharged condition.

Table two:

t	Idm	Vdm	Dd	Testing result
					505～623	0.3316	0.266	0.0195	Abnormal
664～671	0.0294	0.2834	0.00000478	Normally
					997～1014	0.1503	0.3148	0.0038	Abnormal

Wherein, the 1st discharge time section be the same day the 505th～623 point, K nearest neighbor distance and be 0.0195 more than putting Distance threshold (T under electricity condition_d2=0.000494), thus testing result is electric discharge of corresponding moment dystropy；During the 2nd electric discharge Between section be the same day the 664th～671 point, K nearest neighbor distance and be 0.00000478, it is less than the distance threshold under discharge condition Value, thus testing result is that corresponding moment electric discharge behavior is normal；3rd discharge time section be the same day the 997th～1014 point, K Nearest neighbor distance and be 0.0038, it is more than the distance threshold under discharge condition, thus testing result be that corresponding moment electric discharge is capable For exception.

The data in other each moment can be processed in the same way, obtain corresponding energy-storage battery charge and discharge behavior Abnormal or normal testing result.Gained charge and discharge unusual checking result shows, the embodiment of the present invention reflects effectively The ANOMALOUS VARIATIONS of energy-storage battery charge and discharge behavior, has stronger practical significance, for assessing the charging-discharging performances of energy-storage battery Provide the foundation.

Although in above-described embodiment, each step is described according to the mode of above-mentioned precedence, but this area Those of skill will appreciate that, in order to realize the effect of the present embodiment, perform not necessarily in such order between different steps, It can simultaneously (parallel) perform or perform with reverse order, these simply change all protection scope of the present invention it In.

Based on identical with said method embodiment technology design, the embodiment of the present invention also propose a kind of energy-storage battery fill/ Electric discharge unusual checking system.As shown in Figure 4, this detecting system may include that extraction module 42, state determining module 44, Distance and determine module 46 and detection module 48.Wherein, extraction module 42 is for extracting energy-storage battery under charge/discharge state Electric current, voltage data.State determining module 44 for the current data that will extract with charge/discharge current threshold ratio relatively, and according to Comparative result determines the described charge/discharge state of energy-storage battery.Distance and determine that module 46 is for utilizing K arest neighbors method, really Determine in energy-storage battery current-voltage spatial correspondence under charge/discharge state K nearest neighbor distance and.Detection module 48 For by K nearest neighbor distance under charge/discharge state and comparing with the distance threshold under charge/discharge state, and according to Comparative result detects the charge/discharge Deviant Behavior of energy-storage battery.

The module of above-described embodiment can merge into a module, it is also possible to is further split into multiple submodule.

It should be understood that the quantity of the modules in Fig. 4 is only schematically.According to actual needs, each module is permissible There is arbitrary quantity.

Said system embodiment may be used for performing said method embodiment, its know-why, is solved the technical problem that And the technique effect of generation is similar, person of ordinary skill in the field is it can be understood that arrive, for the convenience described and letter Clean, the specific works process of the system of foregoing description and relevant explanation, it is referred to the corresponding process in preceding method embodiment, Do not repeat them here.

Particular embodiments described above, has been carried out the purpose of the present invention, technical scheme and beneficial effect the most in detail Describe in detail bright it should be understood that the foregoing is only the specific embodiment of the present invention, be not limited to the present invention, all Within the spirit and principles in the present invention, any modification, equivalent substitution and improvement etc. done, should be included in the protection of the present invention Within the scope of.

Claims (7)

1. an energy-storage battery charge/discharge anomaly detection method, it is characterised in that described detection method includes:

Extract described energy-storage battery electric current under charge/discharge state, voltage data；

By the current data and the charge/discharge current threshold ratio that extract relatively, and determine the institute of described energy-storage battery according to comparative result State charge/discharge state；

Utilize K arest neighbors method, determine that described energy-storage battery current-voltage space correspondence under described charge/discharge state is closed In system K nearest neighbor distance and；

By K nearest neighbor distance under described charge/discharge state and compare with the distance threshold under charge/discharge state, and The charge/discharge Deviant Behavior of described energy-storage battery is detected according to comparative result.

Detection method the most according to claim 1, it is characterised in that the described energy-storage battery of described extraction is at charge/discharge shape Electric current under state, voltage data, specifically include:

Based on training set, build current data rectangular histogram；

Based on described current data rectangular histogram, by CURRENT DISTRIBUTION probability threshold value calculating current frequency threshold value and according to described electric current frequency Subthreshold determines the off working state of described energy-storage battery；

Characteristic distributions based on current data, rejects electric current occurrence number more than the electricity under the described off working state of pre-determined number Stream, voltage data, the electric current after being screened, voltage data.

Detection method the most according to claim 1, it is characterised in that described charge/discharge current threshold value is according in the following manner Determine:

Judge that whether the current data of described extraction is more than zero；

If it is, determine that described energy-storage battery is in described discharge condition, and determine described discharge current according to below equation Threshold value:

I_Td=min (I_d)

Wherein, described I_dRepresent the current data of described extraction；Described I_TdRepresent described discharging current threshold；

Otherwise, it determines described energy-storage battery is in described charged state, and determine described charging current threshold value according to below equation:

I_Tc=max (I_c)

Wherein, described I_cRepresent the current data of described extraction；Described I_TcRepresent described charging current threshold value.

Detection method the most according to claim 1, it is characterised in that the distance threshold under described charge/discharge state according to In the following manner determines:

The described current-voltage spatial correspondence under described charged state based on training set structure is calculated according to below equation In, the often electric current under group trickle charge state, voltage data average in its charged state persistent period section:

$I_{cm}=\frac{1}{s}\underset{i=1}{\overset{s}{\Σ}}{I}_{c,t_i}$

$V_{cm}=\frac{1}{s}\underset{i=1}{\overset{s}{\Σ}}{V}_{c,t_i}$

Wherein, described s represents the charged state persistent period；Described t_iRepresent the moment that charged state is corresponding；DescribedDescribedRepresent t under current trickle charge state respectively_iThe electric current in moment, voltage data value；Described I_cm, described V_cmRepresent respectively Electric current under described often group trickle charge state, voltage data average in its charged state persistent period section；

The electric current under all continuous discharge states, voltage data is calculated in its discharge condition persistent period section according to below equation Average:

$I_{dm}=\frac{1}{s^{\′}}\underset{i=1}{\overset{s^{\′}}{\Σ}}{I}_{d,t_i}$

$V_{dm}=\frac{1}{s^{\′}}\underset{i=1}{\overset{s^{\′}}{\Σ}}{V}_{d,t_i}$

Wherein, described s ' represents the discharge condition persistent period；Described t_iRepresent the moment that described discharge condition is corresponding；Described DescribedRepresent t under current continuous discharge state respectively_iThe electric current in moment, voltage data value；Described I_dm, described V_dmRespectively Represent the electric current under described all continuous discharge states, the voltage data average in its discharge condition persistent period section；

Electric current, voltage data under each described trickle charge, discharge condition is calculated in described training set respectively according to below equation K nearest neighbor distance of average and:

$D_{c,j}=\underset{i=1}{\overset{K}{\Σ}}\left|\right|\left(\begin{array}{c}{I}_{cm,j}\\ V_{cm,j}\end{array}\right)-\left(\begin{array}{c}{I}_{cm,i}^{*}\\ V_{cm,i}^{*}\end{array}\right)|{|}_2^2$

$D_{d,j}=\underset{i=1}{\overset{K}{\Σ}}\left|\right|\left(\begin{array}{c}{I}_{dm,j}\\ V_{dm,j}\end{array}\right)-\left(\begin{array}{c}{I}_{dm,i}^{*}\\ V_{dm,i}^{*}\end{array}\right)|{|}_2^2$

Wherein, described I_cm,jRepresent current data average under jth trickle charge state；Described V_cm,jRepresent that described jth is continuous Voltage data average under charged state；DescribedDescribedRepresent jth electric current, voltage under described charged state respectively The corresponding electric current of i-th distance, voltage data average in K arest neighbors of data mean value；Described D_c,jRepresent that described jth is even Electric current, voltage data average I under continuous charged state_cm,j、V_cm,jK nearest neighbor distance and；Described I_dm,jRepresent that jth is continuous Current data average under discharge condition；Described V_dm,jRepresent voltage data average under described jth continuous discharge state；DescribedDescribedRepresent in K arest neighbors of jth electric current under described discharge condition, voltage data average corresponding the respectively The electric current of i distance, voltage data average；Described D_d,jRepresent that electric current under described jth continuous discharge state, voltage data are equal Value I_dm,j、V_dm,jK nearest neighbor distance and；Described K takes odd number；Described j span is each filled continuously by described training set Electricity, discharge condition number determine；

Under described charge/discharge state, according to predetermined confidence level α by all described electric currents, voltage data average K Nearest neighbor distance with distance corresponding to α quantile be defined as the distance threshold under charge/discharge state.

Detection method the most according to claim 1, it is characterised in that described method also includes:

Obtain the original electric current of described energy-storage battery, voltage data；

Traversal total data, and with spot patch foot abnormal storage point, it is thus achieved that electric current per minute, voltage data；

It is normalized according to below equation:

$I_n=\frac{I}{100}$

V_n=V-3

Wherein, described I represents described current data per minute；Described V represents described voltage data per minute；Described I_nTable Show the current data after normalization；Described V_nRepresent the voltage data after normalization.

Detection method the most according to claim 5, it is characterised in that described method also includes: carry out parameter initialization, its In, parameter includes obtaining described primary current, the natural law of voltage data, training set size, distribution probability threshold value, K arest neighbors side K value and confidence level in method.

7. an energy-storage battery charge/discharge unusual checking system, it is characterised in that described detecting system includes:

Extraction module, for extracting described energy-storage battery electric current under charge/discharge state, voltage data；

State determining module, for current data and the charge/discharge current threshold ratio that will extract relatively, and determines according to comparative result The described charge/discharge state of described energy-storage battery；

Distance and determine module, is used for utilizing K arest neighbors method, determines that described energy-storage battery is under described charge/discharge state In current-voltage spatial correspondence K nearest neighbor distance and；

Detection module, for by K nearest neighbor distance under described charge/discharge state and with the distance threshold under charge/discharge state Value compares, and detects the charge/discharge Deviant Behavior of described energy-storage battery according to comparative result.